Are engineered nano iron oxide particles safe? an environmental risk assessment by probabilistic exposure, effects and risk modeling

Transcription

1 ISSN: (print), (electronic) Nanotoxicology, 2016; 10(10): ! 2016 Informa UK Limited, trading as Taylor & Francis Group. DOI: / ORIGINAL ARTICLE Are engineered nano iron oxide particles safe? an environmental risk assessment by probabilistic exposure, effects and risk modeling Yan Wang, Lei Deng, Alejandro Caballero-Guzman, and Bernd Nowack Empa, Swiss Federal Laboratories for Materials Science and Technology, Technology and Society Laboratory, St. Gallen, Switzerland Abstract Nano iron oxide particles are beneficial to our daily lives through their use in paints, construction materials, biomedical imaging and other industrial fields. However, little is known about the possible risks associated with the current exposure level of engineered nano iron oxides (nano-feox) to organisms in the environment. The goal of this study was to predict the release of nano-feox to the environment and assess their risks for surface waters in the EU and Switzerland. The material flows of nano-feox to technical compartments (waste incineration and waste water treatment plants) and to the environment were calculated with a probabilistic modeling approach. The mean value of the predicted environmental concentrations (PECs) of nano-feox in surface waters in the EU for a worst-case scenario (no particle sedimentation) was estimated to be 28 ng/l. Using a probabilistic species sensitivity distribution, the predicted noeffect concentration (PNEC) was determined from ecotoxicological data. The risk characterization ratio, calculated by dividing the PEC by PNEC values, was used to characterize the risks. The mean risk characterization ratio was predicted to be several orders of magnitude smaller than 1 ( ). Therefore, this modeling effort indicates that only a very limited risk is posed by the current release level of nano-feox to organisms in surface waters. However, a better understanding of the hazards of nano-feox to the organisms in other ecosystems (such as sediment) needs to be assessed to determine the overall risk of these particles to the environment. Keywords Environmental risk assessment, ecotoxicology, risk assessment, exposure modeling, risk modeling History Received 22 June 2016 Revised 5 August 2016 Accepted 26 September 2016 Published online 10 November 2016 Introduction Assessing the environmental risks of engineered nanomaterials (ENMs) has received a lot of attention in the last years. Whereas for ENMs such as nano-ag and nano-tio 2, a large amount of research on exposure, hazard and risk has been published, other materials have only received very limited attention despite potentially high production amounts. An example of these under-studied materials is engineered nano-iron oxides (Nano- FeOX). Iron oxides are not only engineered materials, but are also present in various forms in natural systems, both in bulk as well as in nano-form (Cornell & Schwertmann, 2003). Nano-scale iron oxides can be formed under mineral weathering process and biotic processes or nanobiomineralization (Hochella et al., 2008; Konishi et al., 2012; Navrotsky et al., 2008). Naturally occuring nano-scaled iron oxides are widely distributed throughout the atmosphere, ocean, water and most living organisms (Guo & Barnard, 2013). Engineered nano-feox has unique properties, which can be applied in various industrial fields. Nano-FeOX can be fully dispersed in a matrix due to the small size and do not scatter light anymore. Therefore, they can be used to manufacture transparent FeOX coatings with high UV absorption and increased durability (Sreeram et al., 2006). FeOX based materials have been found to Correspondence: Bernd Nowack Empa, Swiss Federal Laboratories for Materials Science and Technology, Technology and Society Laboratory, Lerchenfeldstrasse 5, St. Gallen 9014, Switzerland. nowack@empa.ch be effective catalysts involved in oxygen evolution processes, such as water splitting and chlorine evolution (Miyata et al., 1978; Pereira et al., 2012; Rajabi et al., 2013). Nano-sized FeOX particles are more efficient in catalysis than micro-sized particles (Mohapatra & Anand, 2010). Magnetic nano-feox can be used in both in vivo and in vitro biomedical application. e.g. drug targeting, hyperthermia and nuclear magnetic resonance imaging, (Blazkova et al., 2013; Bonnemain, 1998; Cengelli et al., 2009; Tran & Webster, 2010). With their large surface area and adjustable surface charge, nano-feox can be used to absorb cations such as Cd, Co, Zn etc. and anions such as PO 4 3 (H.B. Guo & Barnard, 2013; Luengo et al., 2006; Sumner, 1963). Moreover, when the particle size is below a few nanometers, nano-feox become superparamagnetic, which facilitates phase separation processes after metal binding and selective adsorption (Uheida et al., 2006a,b). Due to these diverse uses and an increasing market demand, nano-feox can be expected to be released to the environment. However, little is known regarding the environmental exposure and risk of nano-feox. It is therefore imperative to quantify the release of nano-feox to natural systems and determine their possible risks to organisms. To assess the release of nano-feox to the environment from applications, a modeling approach is currently the only choice because analytical methods are not advanced enough to distinguish between engineered and naturally occurring nanoparticles (Nowack et al., 2015). Exposure models for ENMs have evolved since Mueller and Nowack initially set the scene with a first

2 1546 Y. Wang et al. Nanotoxicology, 2016; 10(10): material flow modeling framework (Mueller & Nowack, 2008). The model was updated with a lifecycle concept to trace ENMs from cradle to grave and a probabilistic approach to address the uncertainty of the input parameters (Gottschalk et al., 2010; Gottschalk et al., 2009). Material flow models have been applied to calculate the mass flows and concentrations for various ENMs in different regions (Gottschalk et al., 2015; Mahapatra et al., 2015; Sun et al., 2014; Wang et al., 2016). Keller et al. (2013) predicted the combined flows of nano-feox/zero-valent iron (NZVI) to landfill, air, water and soil at a global level, where seven industrial sectors were considered (Keller et al., 2013). In a follow-up study in 2014, the same authors calculated the released amount of nano-feox/nzvi for eight different regions and predicted the concentration of nano-feox/nzvi in wastewater treatment plant effluents and bio-solids in California (Keller & Lazareva, 2014). The models in the two aforementioned studies used mathematic algorithms, in which the input parameters are based on data from one market research report (production volume and distribution to products of nano-feox/nzvi). The toxic effects of nano-feox have been studied including short-term and long-term toxicity (Arami et al., 2015; Hazeem et al., 2015; Mahmoudi et al., 2011; Mahmoudi et al., 2009; Weissleder et al., 1989; Zhang et al., 2015; Zhu et al., 2012). Early toxicity studies for nano-feox found that their toxicity is low and that there is no link between size and toxicity (Karlsson et al., 2009; Kunzmann et al., 2011). However, some later studies found that nano-feox can be cytotoxic and can induce systemic toxicity (Mahmoudi et al., 2010; Zhu et al., 2012). Prodan et al. (2013) have shown that nano-feox can be toxic to rats and a sizedependent toxicity has been found by Hazeem et al. (2015). Zhang et al. (2015) observed a high accumulation of nano-feox in Zebrafish, accompanied, however, by a high elimination rate. In environmental risk assessment, the risk is quantified by comparing the predicted environmental concentration (PEC) and the predicted no effect concentration (PNEC) (ECHA, 2008b). The PNEC represents the critical concentration at which no effects are expected in a given ecosystem. The PNEC can be obtained from the lowest observed no-effect concentration (NOEC) or from statistical extrapolation methods using species sensitivity distribution (SSD) (ECHA, 2008b). SSDs have been built for various ENMs (Garner et al., 2015; Gottschalk & Nowack, 2013; Semenzin et al., 2015). However, up to date, no evaluation using SSD has been performed so far for nano-feox. NZVI and nano-feox have common properties (e.g. magnetic properties) and similar potential for environmental applications and therefore they are often grouped together, e.g. in Keller et al. (2013). NZVI has emerged as a new option in treating soil and groundwater with contaminants (Mueller et al., 2012). Due to its susceptibility to oxidation, it is rapidly oxidized to form an iron oxide when oxygen is present. To estimate the environmental risk of nano-feox more precisely, in the current paper the target material is nano-feox only, not including NZVI or its transformation products. To date, no complete environmental risk assessment for nano- FeOX has been conducted. Therefore, this study aims to assess the release of nano-feox to the EU and Swiss environment, predict exposure levels and assess the risks of nano-feox to the environment using a probabilistic material flow modeling and risk assessment approach. Methods Model description A material flow modeling approach was used to predict the mass flows and environmental concentrations of nano-feox for the targeted environmental compartments, including air, water, soil, and sediments as shown in Figure 1. The targeted system boundaries in this study are EU and Switzerland (CH). The model is based on a material flow analysis from a life cycle perspective to analyze mass flows of nano-feox to the environment as it can be released through the whole lifetime (Gottschalk et al., 2009; Sun et al., 2014; Wang et al., 2016). The model includes four different stages: production of nanoparticles; manufacturing and consumption of consumer products incorporating nano-feox; waste management processes; and environmental compartments. Mass loss during production and manufacturing processes were considered in the model, as described in Mueller and Nowack (2008). Waste management compartments include landfill, waste incineration plant (WIP), wastewater treatment plant (WWTP), recycling and export. Potential leaching from landfills was not included in this iteration of the model. However, the behavior of nano-feox after it flows into the recycling processes was Figure 1. Framework of the material flow model. The framework contains four parts: production; manufacturing & consumption; waste management; and the environment. Waste management includes landfills, waste incineration plants (WIP), waste water treatment plants (WWTP), recycling and export. There are four environmental compartments: air, water, soil and sediment. The arrow from the waste management box to manufacturing represents reuse of materials after recycling.

3 DOI: / Environmental risk assessment of nano iron oxides 1547 considered based on the study from Caballero-Guzman et al. (2015). However, this investigation was only done for four product categories: Electronics ; Paints ; Battery ; and Construction because no data were reported for the other product categories in (Caballero-Guzman et al., 2015). As the basis for the previous recycling model only targeted the Swiss system, we extrapolated this data and assumed that the fate of nano-feox during recycling is the same in the EU as in Switzerland. Environmental compartments considered in the modeling were air, soil, surface water, and sediment. In the case of the EU, the soil compartment was separated into natural & urban soil (N&U soil) and sludge treated soil (ST soil) because sewage sludge is applied as fertilizer in most European countries. The retention time of nanoparticles in the air and surface water system were defined as 10 and 40 days, respectively (Anastasio & Martin, 2001; ECB, 2003). Two worst-case scenarios were chosen for the aquatic system: no sedimentation for the surface water compartment and complete sedimentation for the sediment compartment. It was assumed that all compartments are homogenous and wellmixed systems. Based on the model framework, three fundamental input parameters are required: production volume, product allocation and transfer factors among different compartments. Input parameters Production volume The goal was to predict the material flows and risk of nano-feox in the environment in the calendar year of The most fundamental input parameter of the model is the production volume. Data regarding produced amounts of nano-feox was collected from different sources such as market research reports and official registration reports, reflecting the production values for the year 2014 (for more information, see SI). All production data of years other than 2014 were projected to 2014 based on a published general development of ENMs production over time (Sun et al., 2016). We assumed that the development of production volume of nano-feox was proportional to the general development of ENMs production as no specific data for nano- FeOX are available. The ratios between each year to the reference year 2014 were brought into the model as a probability distribution, presenting a comprehensive picture of the real production volume for the calculated year compared to the reference year All data were extrapolated to estimate the production volume in the EU and Switzerland using GDP ppp (Gross Domestic Product per capita based on Purchasing Power Parity) (Wang et al., 2016). To handle the uncertainty in production volumes from various datasets, each value was introduced to the model as a probability distribution (Sun et al., 2014; Wang et al., 2016). Moreover, a Degree of Belief (DoB), 20 or 80, was assigned to each source as a measure of reliability. The detailed explanation and the decision tree of choosing a certain DoB is described in the supporting information. The DoB of a data source was reflected by the sample size during the modeling. The sample size of each input parameter is 100,000. Hence, the sampling sizes were for example 20,000 for data sources with DoB of 20 and 80,000 for those with DoB of 80. Product allocation The distribution of nano-feox to nano-products is the second most important input parameter for our modeled system. This allocation determines the further fate of nano-feox based on the life cycles of different product categories. Because no single database provides a comprehensive overview of products containing nano-feox, a list of categories for nano-feox was defined based on market reports and the official French registration report (ANSES, 2013; Future Markets, 2014; SRI, 2011). Four different methods were used to generate the share of nano- FeOX in different categories: M1- using the data in the Future Markets report from 2012 to calculate the allocation of nano- FeOX in each category (Future Markets, 2012); M2- a patent search of U.S. and European patent inventories (EPR, 2015; USP&TO, 2015); M3- using the known market value of each product category incorporating nano-feox to calculate the product distribution; and M4- taking available direct production data for nano-feox in a product category. A DoB was assigned to each method depending on their reliability. A detailed description of the four methods and the data collected is described in the Table S2 of the supporting information. Transfer factors Transfer factors determine how much nano-feox flows between different compartments of the model. However, there is no study available regarding releases of nano-feox from commercial applications. The transfer factors were therefore based on those of the same product categories containing engineered nano-tio 2 from Sun et al. (2014). This is because nano-feox is used in the same or similar products as engineered nano-tio 2 such as paints, electronics and cosmetics, which have the same or similar lifecycles. Table S3 presents where nano-feox goes during use and disposal. WWTP and WIP are technical processes where significant amounts of nano-feox are captured. Therefore, the removal efficiencies of these processes are critical for the environmental releases of nano-feox. Nano-FeOX share similar properties with other nano metal oxides e.g. solubility and reactivity. No information is available regarding removal efficiency of nano-feox during wastewater treatment or WIP. We therefore used the removal rates of total iron and nano-tio 2 (Table S4) to build a probability distribution (Figure S2) for nano-feox. For the removal rate in WIP, the data for engineered nano-ceo 2 from a full-scale WIP was chosen (Walser et al., 2012). The transfer factors within the recycling systems are chosen from Caballero- Guzman et al. (2015). Hazard assessment For the hazard assessment, the potential hazardous effects posed by nano-feox to organisms were evaluated and PNECs for a specific environmental system were obtained. The collected endpoint concentrations (EC50, EC10, EC55, LC50, NOEC and LOEC) contain ecotoxicological data of both modified nano-feox (e.g. coated nano-feox) and nonmodified nano-feox (pure nano-feox particles). All sensitivity concentrations were converted to NOEC values using assessment factors. The criteria for choosing the assessment factors were described in our previous study (Wang et al., 2016). We generated the PNEC values from the 5th percentile of cumulative SSD as recommended by the REACH guidance (ECHA, 2008a). In this study, we combined individual probabilistic species sensitivity distributions (PSSD) of each species to generate one PSSD (Gottschalk & Nowack, 2013; Wang et al., 2016). With simulation runs, a PNEC distribution was derived by abstracting the 5th percentiles from PSSDs (Coll et al., 2016). Due to the limited availability of ecotoxicological data, the risk assessment for nano-feox is currently only feasible for surface waters. No data were found for sediments and for soil organisms only results from seed germination tests on filter paper are available but not for real soils.

4 1548 Y. Wang et al. Nanotoxicology, 2016; 10(10): Risk assessment The environmental risk was quantified by comparing the PEC with the PNEC values. The risk characterization ratio (RCR), the ratio of PEC over PNEC shown in Equation 1, was used as a measure of the risk level [52]. Here, if the RCR value is over 1, it indicates that the risk management measures should be taken regarding a given material. If not, it suggests the risk is under control at the current release level. RCR ¼ PEC=PNEC Instead of calculating one single RCR, a RCR distribution was derived by dividing all PEC values by all PNEC values of the respective probability distributions (Coll et al., 2016). Results Exposure assessment Nano-FeO x production values The data sources used to obtain production volume of nano-feox with their assigned DoBs are shown in Table S1. The values are valid for different years and are extrapolated to 2014 using the scaling factors from Sun et al. (2016). The summary of the production volumes for the EU and Switzerland in 2014 is presented in Table 1. The modeled mean production amount is 7500 tons/year for the EU, with a range of the 15 85% percentiles from 2.9 to 9400 tons/year. The reported production numbers cover a very large range, indicating a high uncertainty about the amount produced. The scaled mean production for Switzerland is 190 tons/year. Because no information shows if the produced materials are all used up within the reference year or stored for next year onwards, we equate production to use as done in all previous studies about material flows of ENMs. Product allocation A total of 14 product categories were identified based on the information collected using the four different methods. Table S2 shows the share of nano-feox in each product category calculated by the different methods, as well as the assigned DoB of each method. The weighted mean values of the shares of nano-feox in each category from all methods were calculated using the DoB and they were included in the model as triangular distributions. The nine major categories with the share of nano-feox used in each category are displayed in Figure 2. The category Others expresses those categories where less than 1% of nano-feox is used. 37.6% of total nano-feox is applied in paints followed by the categories Electronics and Construction, which account for 16 and 12.9%, respectively. These estimations are consistent with the Future Markets report (2014), which is primarily based on interviews. It was reported that high volume applications are pigments, polishing and electronics. A significant amount of nano- FeOX is allocated to the categories Cosmetics, Catalyst and Medical. Smaller fractions of nano-feox are used in batteries, water treatment and plastics. ð1þ removal efficiency during wastewater treatment are shown in Table S4, the resulting probability distribution is the transfer factor in Figure S2 of the Supporting Information. Based on the input parameters mentioned above, probability distributions for each flow were generated. The median, mean, Q15 and Q85 values were abstracted from each distribution and are given in Table S5. Figure 3 illustrates the mean mass flows, giving an overview of the mass distribution of nano-feox among all compartments in the EU and Switzerland. The only inflow is from production, manufacturing and consumption. Nano- FeOX flows out of the system boundary through export or recycling. This is because some wastes in the EU and Switzerland are exported to other counties to be reused or recycled (e.g. E-waste). The most dominant flow is from Production, manufacturing and consumption (PMC) to recycling tons nano-feox in the EU goes to recycling processes and this number for Switzerland is 75 tons. Another significant flow is from PMC to landfill, in the EU, 1700 out of total 7500 tons of produced nano-feox (75 out of 190 tons in Switzerland). The main flows to surface water are from the PMC compartment, the untreated wastewater and from the effluent from wastewater treatment plants. The amount of nano- FeOX released to the air is very small and is mostly from the PMC compartment. The nanoparticles in air further deposit to the soil and surface water. Soils and sediment were compartments where nano-feox is likely to accumulate. The main input to soil is through application of sewage sludge. However, in Switzerland, sludge is incinerated or used as alternative fuel in cement plants and therefore the overall flow of nano-feox to soil is much smaller than in the EU. The fate of nano-feox during recycling processes was also included in the model (Table S5). After entering the recycling process, nano-feox flows to Cement, WIP, Export, PMC and Landfill. Nano-FeO x flows into the recycling compartment from nine product categories, but due to limited data availability we could only include the further fate of nano- FeOX contained in Paints, Construction, Electronics and Battery (Caballero-Guzman et al., 2015). A significant percentage of nano-feox (2600 out of 3100 tons in the EU, 64 out of 75 tons in Switzerland) went to recycling and could be further allocated to outgoing flows. The remaining mass was attributed to the unconsidered sub-compartment and not Mass flows of nano-feox Before calculating the mass flow to different compartments, transfer factors were assessed (Table S3). The data collected for Table 1. Summary of production volumes in the EU and Switzerland in 2014 (Tons). System Q15 Mean Median Q85 EU CH Figure 2. Distribution of nano-feox in product categories.

5 DOI: / Environmental risk assessment of nano iron oxides 1549 Figure 3. Material flows of nano-feox in the EU (top) and CH (bottom) in tons/year. Nano-FeOX flows out of the same compartment are shown in the same color. All values are rounded to two significant numbers. The input and output flows are balanced in terms of mean values of the probabilistic distribution. Unconsidered represents product categories, which were not considered in the recycling process evaluation of Caballero- Guzman et al. (2015). analyzed any further. A major fraction of the recycled materials containing nano-feox was landfilled (1700 tons in the EU and 42 tons in Switzerland). In addition, a noticeable flow from recycling processes goes to waste incineration plants (440 out of 2600 tons in the EU). It is obvious from the flow charts that differences in the waste management process exist between the EU and Switzerland that have a profound influence on the final mass flows, e.g. sludge from the WWTP is incinerated in Switzerland (78%), while it can be landfilled (20%) or applied as fertilizer in the EU (55%). Predicted nano-feox concentrations in technical and environmental compartments The predicted nano-feox concentrations were calculated by dividing the mass flows by the mass or volume of each compartment, as described in further detail in Sun et al. (2014). These data therefore provide average concentrations in wellmixed systems. Environmental concentrations were estimated for four environment compartments: air, soil (natural and urban soil (N&U soil) and sludge-treated soil (ST soil)), surface water and sediment; in addition, concentrations of nano-feox in effluent

6 1550 Y. Wang et al. Nanotoxicology, 2016; 10(10): Table 2. Concentrations of nano-feox in technical and environmental compartments. Q15 Mean Median Q85 Unit EU Air 1.1E E E E-01 ng/m 3 Sediment 1.6E mg/kgy N&U a soil 3.8E E E E-01 mg/kgy ST a soil 4.4E mg/kgy Surface water 1.1E E ng/l Waste 1.2E mg/kg WIP a bottom ash 7.0E mg/kg WIP a fly ash 1.0E mg/kg WWTP a effluent 4.9E E mg/l WWTP a sludge 3.1E E mg/kg CH Air 2.7E E E E-01 ng/m 3 Sediment 1.5E mg/kgy N&U a soil 9.2E E E E-01 mg/kgy ST a soil mg/kgy Surface water 2.0E E ng/l Waste 2.4E mg/kg WIP a bottom ash 4.0E mg/kg WIP a fly ash 5.0E mg/kg WWTP a effluent 6.6E E mg/l WWTP a sludge 3.7E mg/kg For soil and sediments the yearly increase in concentration is given. All values are rounded to two significant numbers. a N&U soil: Natural and urban soil; ST soil: Sludge treated soil; WIP: Waste incineration plant;. WWTP: Wastewater treatment plant. Table 3. Test species and taxonomic groups, and total number of endpoints. Taxonomic group Test organism Total number of endpoints available Unicellular V. fischeri, E. coli, P. phosphoreum, R. subcapita 9 Invertebrate D. magna, B. plicatilis, P. subcapita, C. dubia, D. magna, T. thermophila 11 Vertebrate D. rerio, O. latipes, 6 and sludge from wastewater treatment plants were calculated, as well as concentration of nano-feox in the waste, WIP fly ash, and WIP bottom ash. Table 2 presents the mean, median, Q15 and Q85 values of the predicted concentrations. The concentrations for soils and sediments correspond to the yearly increases in concentrations since the ENMs are assumed to accumulate there. The lowest concentration is found in air (EU, 0.29 ng/m 3 and CH, 0.76 ng/m 3 ). This is because the application of nano-feox in products results only in small amount of releases to the atmosphere. In the complete sedimentation scenario, the concentration in sediments is predicted to be the highest of all environmental concentrations (EU, 390 mg/kgy and CH, 450 mg/ kgy). However, our model is a static model and therefore the accumulation of nano-feox in all compartments from previous years was not considered. The calculated concentrations are only based on the input flow in Hazard assessment The hazard assessment using SSDs could only be conducted for the surface water compartment because insufficient ecotoxicity studies with usable endpoint concentrations for soils and sediments were available. From freshwater studies, 26 endpoint concentrations were collected from 12 species from unicellular, invertebrate and vertebrate groups (Table 3). The original test concentration, assessment factors and the calculated NOEC values are displayed in Table S6 in the supporting information. Figure 4. Cumulative PSSD for nano iron oxides in the surface water compartment: the red curve is the cumulative PSSD; green dots are the NOEC values. Figure 4 shows the NOEC values considered in this work and the cumulative SSD. The most sensitive species from the data points collected is a unicellular organism, T. thermophila (Tetrahymena thermophila). C. dubia (Ceriodaphnia dubia) is the least affected species by nano-feox in surface water from the

7 DOI: / Environmental risk assessment of nano iron oxides 1551 Figure 5. The predicted no-effect concentration (PNEC) distribution (in red) compared with the probability density distribution of predicted environmental concentrations (PEC) (in blue) for nano-feox in the European surface water system.. species studied thus far. The NOEC values of Daphnia magna differ by two orders of magnitude (from 23 mg/l to 5000 mg/l), a range that is smaller than that for other ENMs (where the data for one species can have a range of up to five orders of magnitude) (Coll et al., 2016). The probability distribution of the PNEC was built based on the 5th quantile from each PSSD generated during the simulation run. The obtained mean value of the PNEC is 218 mg/l, with a range from 169 mg/l (Q15) to 267 mg/l (Q85). A sensitivity analysis was conducted to evaluate the influence of the lowest endpoint concentration to D. magna on the PNEC value. A new PSSD, obtained by removing the lowest NOEC value for D. magna is given in Figure S3 in the Supporting Information. The new PNEC value was calculated to be 239 mg/l, an increase by 8% compared to the PNEC from the complete data set. Risk characterization The distributions of PEC and PNEC are plotted together in Figure 5 for the EU. The corresponding plot for Switzerland is shown in Figure S4. This plot shows that the expected environmental concentrations are several orders of magnitude below the concentrations where effects can be expected. The risk characterization ratios were calculated by dividing all values of the PEC distribution by those of the PNEC distribution generated from the complete collected data points. The mean values of the RCR are for the EU and for Switzerland. The likely range of the RCR is from (Q15) to (Q85) for the EU (for CH from to ), which indicates that the potential risk posed by nano-feox in the EU and Swiss surface water systems is very limited. The complete RCR distribution curves are given in Figure S5. Discussion This work provides the first probabilistic flow assessment and the first risk assessment for nano-feox in aquatic systems. Many steps in this risk assessment process are based on data with high variability or uncertainty and we discuss in the following the various parts of the modeling in this context. In the current situation with limited knowledge about the fate of ENMs applied in commercial products, all estimations of environmental flows and risks are definitely associated with uncertainties and variability. One of the main causes of uncertainty in the flow modeling presented here is the overall production amount of nano-feox. Data regarding production collected from different sources varies widely. The highest production volumes were obtained from Schmid and Riediker (2008) (scaled to the EU and CH: 32,000 and 365 tons/a), which were about 10,000 and 5000 times more than the production volumes provided by the Future Markets report (2014) (scaled to the EU and CH: 2.7 and tons/a). The Schmid and Riediker value was derived from a mandatory survey of all companies in Switzerland using nanomaterials and therefore constituted a source with high credibility. The Future Markets values are the most recent and come from a source with an in-depth evaluation of production and use of nanomaterials. The variability of the production volume is likely caused by different perspectives of defining a nanomaterial. EU member states, agencies and economic operators are recommended to use the nanomaterial definition provided by the EU Commission (EUCommission, 2011). Studies using other or not clear definitions before 2011 may consider different types of materials as nanomaterials (e.g. including or excluding some traditional materials with similar size distribution). Older data were therefore assigned a lower DoB and thus had less influence on the final result than more recent data. In all mass flow analyses for ENMs that have been performed so far, product allocation is the weakest step in the modeling process, potentially having a large influence on the final mass flows because the life cycles of products are decisive in determining the potential and the magnitude of release. Without exception, another origin of uncertainty of the mass flows of nano-feox is the uncertain allocation of the produced nano- FeOX to different product categories. The properties and lifecycles of the products determine where and how much of nano- FeOX is released. In this study, the product categories were defined mainly based on the French registration report, supported by a survey and market research reports (ANSES, 2014; Future Markets, 2014; Sørensen et al., 2015). The share of nano-feox in each category was determined by four different methods. With this approach we aimed to capture all possible products and applications that are on the market. Certain products only appeared in the result of one method whereas others came out in all of the 4 methods. These certain product categories such as paints and electronics also have the major market share and therefore dominated the overall flows of nano-feox. A product category only covered by one evaluation method makes up a small share of the total mass and it only had a small influence on the overall share of nano-feox in this category. The use of different methods therefore increased our confidence in the product allocation obtained. Transfer factors decide where nano-feox distributes after the nano-enabled products were used. The dominant transfer factors are the release fractions from a product to technical and environmental compartments. Other important factors are the removal efficiency during waste management processes. Release from nano-feox enabled products, in most cases, is not nano specific, but depends on the life cycles of the products and we have therefore used in most cases the transfer factors for nano- TiO 2 also for nano-feox. This is justifiable since both engineered nano-tio 2 and nano-feox are rather insoluble and have similar chemical and physical properties. Therefore, it is possible to use previous release assessments of products containing other ENMs for nano-feox-enabled products. Some of transfer factors of nano-feox in this work were based on surveys, e.g. for Paints and Construction (Hincapie et al., 2015). Other transfer factors were chosen from the transfer factors of engineered nano-tio 2 for the same product categories or the categories with similar life cycles from Sun et al. (2014). The information

8 1552 Y. Wang et al. Nanotoxicology, 2016; 10(10): provided by experimental release studies is rather incomplete concerning descriptions and characterizations of ENMs and the released materials (Caballero-Guzman & Nowack, 2016). Limited knowledge is available about the forms of released ENMs and especially data from real-world situation are missing. However, various approaches have been developed to derive transfer factors as described by Caballero-Guzman and Nowack (2016). The flow analysis revealed that nearly 40% of nano-feox was discharged to recycling in both the EU and Switzerland. This can be explained by the fact that a significant amount of nano-feox is used in electronics, paints, and construction, which were recycled in significant amounts as e-waste and construction waste. From production/manufacturing/use, 23% of nano-feox flows to landfill, 13% deposits in wastewater, 15% in waste incineration and 5% of nano-feox flowed out of the system boundaries since the nano-enabled products were exported. The rest of nano-feox was distributed to air, soil, and surface water from PMC. As 40% of nano-feox flows to the recycling system, it was important to further investigate the fate of nano-feox during the recycling processes. Referring to the previous study by Caballero- Guzman et al. (2015), four categories (over 80% of nano-feox) were considered after entering the recycling stage. The majority of the recycled nano-feox was predicted to be landfilled (65%) and WIP (16%), which is a consequence of the use of over 50% of nano-feox in the Paints and Construction categories, which have a high recycling rate of 46 and 70%, respectively (in the EU). Around 8% of recycled nano-feox from the Paints and Construction categories was reused in cement plants and approximately the same amount was reused in other materials coming from construction waste recycling. Nano-FeOX within the product categories Electronics and Battery was also exported as E-waste out of Europe and Switzerland to be reused or recycled. To improve the current model, more studies are needed to investigate specific fate of nano-feox during recycling processes. The distribution of nano-feox to product categories with different life cycles strongly affects the mass of nano-feox flows to environmental compartments, as well as environmental concentrations. The predicted concentrations of nano-feox were calculated to be ng/l in European surface water. The concentration range for engineered nano-tio 2 in Sun et al. (2014) in the EU was estimated to be mg/l, the upper boundary was 38 times larger than that of nano-feox. However, the input flow (production volume) of engineered nano-tio 2 was only 1.5 times larger than that of nano-feox. This is because in the former study, a much higher percentage (60%) of engineered nano-tio 2 is used in cosmetics, which is mainly released to waste water and finally discharged to surface water. There are other modeling studies available for nano-iron (combining both nano-feox and NZVI in the same model) where the authors used the production amount (over 40,000 tons) and the product allocation (seven technological sectors) from one commercially market research report (Keller et al., 2013; Keller & Lazareva, 2014; Liu & Cohen, 2014). We also included the data from this report as one of the sources to define the production amount and as one of the methods to obtain the product allocation (Future Markets, 2012). Both production amount and product distribution are associated with high uncertainty. By combining different sources with different DoB in our approach, we were able to obtain probability distributions and therefore had the possibility to quantify the uncertainty of the model results by providing probability ranges of flows and concentrations. Other differences between the Keller model (Keller et al., 2013; Keller & Lazareva, 2014; Liu & Cohen, 2014) and ours are the selection of product categories and the values for the transfer factors, e.g. the removal efficiencies in WWTP and WIP. Another major difference is that the recycling compartment was not considered in the Keller model. However, this is the compartment where about 40% of nano-feox from PMC flows to in our model and thus represents an important driver of the flows. The PEC values we provide cannot be validated at the moment because analytical measurements are not advanced enough to quantify trace amount of ENMs with a high background concentration of naturally occurring nanoparticles (Nowack et al., 2015). These authors stated that combing modeling with analytical method can provide an overall view of the presence of the nanomaterial in the environment (Nowack et al., 2015). This is especially important for iron oxides which are known to occur naturally in nano-scaled form in all environmental compartments (Sposito, 1989). The riverine suspended sediment concentration was reported up to 50 mg/l (Tockner et al., 2009). The share of particulate iron in suspended sediments in European rivers was investigated by Poulton and Raiswell (2002) and the average riverine particulate iron accounts for 1.4% (wt) (Poulton & Raiswell, 2002). Hence, the average concentration of particulate iron in European rivers can reach up to 70 mg/l. This result is three orders of magnitudes higher than the predicted concentration of nano-feox in this study, indicating that nano-feox particles only have an insignificant contribution to the total concentration of iron in the European environment. The flows of nano-feox in technical compartments, such as waste materials and wastewater treatment plant effluents, can provide input data for environmental fate modeling studies. In addition, the predicted concentrations also can be a reference for policy makers to decide whether it is necessary to monitor an effluent or biosolid (Praetorius et al., 2013) Our work also provides the first hazard assessment for nano- FeOX based on a dataset covering 11 species and 26 endpoint concentrations, but only for aquatic system due to data availability. The lower boundary of the PNEC (169 mg/l) is at least four orders of magnitude larger than the upper boundary of the PEC of nano-feox in surface water (36 ng/l in the EU). Therefore, the current expected amount of nano-feox does not have an adverse impact to organisms in surface waters and the potential risk posed by nano-feox in the surface water is predicted to be limited. There is a high variability associated with the various ecotoxicological endpoint concentrations that we have used to calculate the PNEC. The data cover the toxicity of different iron oxides (nano Fe 3 O 4 and nano Fe 2 O 3 ) but also of functionalized nano Fe 2 O 3 particles. For instance, for D. magna, the smallest observed NOEC is 23 mg/l, which is about 200 times smaller than the largest value (5000 mg/l). Here, the lowest NOEC indicates the toxicity of bare Fe 3 O 4 nanoparticles to D. magna. However, the other five sensitivity concentrations for D. magna correspond to Fe 3 O 4 with four different coatings, indicating that the coating plays a critical role in determining the toxicity of nano iron-oxidebased materials (Baumann et al., 2014). Nano-FeOX is incorporated in products directly or in a modified form. However, there is no data available on production or use of functionalized nano- FeOX and it is therefore not possible to conduct a risk assessment only for modified iron oxides nanoparticles. In this case, to better estimate the hazards of nano-feox applied in products, sensitivity concentrations should cover both the non-modified and the modified form of nano-feox. The probabilistic approach used in this study is able to cope with this uncertainty when defining the PNEC value. A larger set of ecotoxicity endpoints of nano-feox in different forms under various test conditions would help to better determine the hazard of nano-feox. In summary, using the current model framework plus the included applications and transfer factors, less than 15% of produced nano-feox was directly released to environment

9 DOI: / Environmental risk assessment of nano iron oxides 1553 (Air, Soil, Water and sediment). Moreover, the results also indicate that, at the predicted release level, nano-feox does not endanger the aquatic organisms in the EU and Swiss systems. Declaration of interest The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article. This work was supported by the Swiss Federal Office for Environment (BAFU) and a Chinese Government Scholarship to YW. A.C.G. is supported by the project on Sustainable Nanotechnologies (SUN) that receives funding from the European Union Seventh Framework Program [FP7/ ] under Grant Agreement No. [604305]. References Anastasio C, Martin ST Atmospheric nanoparticles. In: Banfield JF, Navrotsky A, eds. Nanoparticles and the Environment. Vol 44. Mineralogical Society of America, ANSES Éléments issus des déclarations des substances à l état nanoparticulaire. RAPPORT d étude. ANSES (l Agence nationale de sécurité sanitaire). ANSES Éléments issus des déclarations des substances à l état nanoparticulaire. RAPPORT d étude. ANSES (l Agence nationale de sécurité sanitaire). Arami H, Khandhar A, Liggitt D, Krishnan KM In vivo delivery, pharmacokinetics, biodistribution and toxicity of iron oxide nanoparticles. Chem Soc Rev 44: Baumann J, Koser J, Arndt D, Filser J The coating makes the difference: acute effects of iron oxide nanoparticles on Daphnia magna. Sci Total Environ 484: Blazkova I, Nguyen HV, Dostalova S, Kopel P, Stanisavljevic M, Vaculovicova M, et al Apoferritin modified magnetic particles as doxorubicin carriers for anticancer drug delivery. Int J Mol Sci 14: Bonnemain B Superparamagnetic agents in magnetic resonance imaging: physicochemical characteristics and clinical applications. A review. J Drug Target 6: Caballero-Guzman A, Nowack B A critical review of engineered nanomaterial release data: are current data useful for material flow modeling? Environ Pollut 213: Caballero-Guzman A, Sun T, Nowack B Flows of engineered nanomaterials through the recycling process in Switzerland. Waste Manag 36: Cengelli F, Grzyb JA, Montoro A, Hofmann H, Hanessian S, Juillerat- Jeanneret L Surface-functionalized ultrasmall superparamagnetic nanoparticles as magnetic delivery vectors for camptothecin. Chemmedchem 4: Coll C, Notter D, Gottschalk F, Sun T, Som C, Nowack B Probabilistic environmental risk assessment of five nanomaterials (nano-tio 2, nano-ag, nano-zno, CNT, and fullerenes). Nanotoxicology 10: Cornell RM, Schwertmann U The Iron Oxides: Structures, Properties, Reactions, Occurences and Uses. Weinheim: Wiley-VCH Verlag GmbH & Co. KGaA. ECB Technical Guidance Document on Risk Assessment, European Chemicals Bureau. ECHA. 2008a. Guidance on information requirements and chemical safety assessment Part B: Hazard assessment. ECHA. 2008b. Guidance on information requirements and chemical safety assessment, Part E risk characterization. EPR European Patent Office. European Patent Register [Online]. Available at: EU Commission Recommendation on the definition of nanomaterial (2011/696/EU). O. J. L 275: 3. Future Markets The global market for nanomaterials : production volumes, revenues and end use markets. Future Markets Inc., [Online]. Available at: com/ index.php?option¼com_content&view¼article&id¼ 176&Itemid¼73. Future Markets The Global Market for Metal Oxide Nanoparticles Future Market. Garner KL, Suh S, Lenihan HS, Keller AA Species sensitivity distributions for engineered nanomaterials. Environ Sci Technol 49: Gottschalk F, Lassen C, Kjoelholt J, Christensen F, Nowack B Modeling flows and concentrations of nine engineered nanomaterials in the Danish environment. Int J Environ Res Public Health 12: Gottschalk F, Nowack B A probabilistic method for species sensitivity distributions taking into account the inherent uncertainty and variability of effects to estimate environmental risk. Integr Environ Assess Manage 9: Gottschalk F, Scholz RW, Nowack B Probabilistic material flow modeling for assessing the environmental exposure to compounds: Methodology and an application to engineered nano-tio 2 particles. Environ Model Software 25: Gottschalk F, Sonderer T, Scholz RW, Nowack B Modeled environmental concentrations of engineered nanomaterials (TiO 2, ZnO, Ag, CNT, fullerenes) for different regions. Environ Sci Technol 43: Guo HB, Barnard AS Naturally occurring iron oxide nanoparticles: morphology, surface chemistry and environmental stability. J Mater Chem A 1:17. Hazeem LJ, Waheed FA, Rashdan S, Bououdina M, Brunet L, Slomianny C, et al Effect of magnetic iron oxide (Fe 3 O 4 ) nanoparticles on the growth and photosynthetic pigment content of Picochlorum sp. Environ Sci Pollut Res Int 22: Hincapie I, Caballero-Guzman A, Hiltbrunner D, Nowack B Use of engineered nanomaterials in the construction industry with specific emphasis on paints and their flows in construction and demolition waste in Switzerland. Waste Manage 43: Hochella MFJ, Lower SK, Maurice PA, Penn RL, Sahai N, Sparks DL, Twining BS Nanominerals, mineral nanoparticles, and Earth systems. Science 319: Tockner K, Uehlinger U, Robinson CT Rivers of Europe. Amsterdam: Academis Press. Karlsson HL, Gustafsson J, Cronholm P, Möller L Size-dependent toxicity of metal oxide particles-a comparison between nano- and micrometer size. Toxicol Lett 188: Keller A, McFerran S, Lazareva A, Suh S Global life cycle releases of engineered nanomaterials. J Nanoparticle Res 15:1 17. Keller AA, Lazareva A Predicted releases of engineered nanomaterials: from global to regional to local. Environ Sci Technol Lett 1: Konishi H, Xu H, Guo HB Nature s Nanostructures. Singapore: Pan Stanford Pte. Ltd. Kunzmann A, Andersson B, Thurnherr T, Krug H, Scheynius A, Fadeel B Toxicology of engineered nanomaterials: focus on biocompatibility, biodistribution and biodegradation. Biochimica Et Biophysica Acta-General Subjects 1810: Liu HH, Cohen Y Multimedia environmental distribution of engineered nanomaterials. Environ Sci Technol 48: Luengo C, Brigante M, Antelo J, Avena M Kinetics of phosphate adsorption on goethite: Comparing batch adsorption and ATR-IR measurements. J Colloid Interface Sci 300: Mahapatra I, Sun TY, Clark JRA, Dobson PJ, Hungerbuehler K, Owen R, Nowack B, Lead J Probabilistic modelling of prospective environmental concentrations of gold nanoparticles from medical applications as a basis for risk assessment. J Nanobiotechnol 13. Mahmoudi M, Laurent S, Shokrgozar MA, Hosseinkhani M Toxicity evaluations of superparamagnetic iron oxide nanoparticles: cell vision versus physicochemical properties of nanoparticles. ACS Nano 5: Mahmoudi M, Simchi A, Imani M, Shokrgozar MA, Milani AS, Hafeli UO, Stroeve P A new approach for the in vitro identification of the cytotoxicity of superparamagnetic iron oxide nanoparticles. Colloids Surf B Biointerfaces 75: Mahmoudi M, Simchi A, Milani AS, Stroeve P Cell toxicity of superparamagnetic iron oxide nanoparticles. J Colloid Interface Sci 336: Miyata T, Ishino Y, Hirashima T Catalytic reduction of aromatic nitro-compounds with hydrazine hydrate in presence of iron(iii) oxide hydroxide. Georg Thieme Verlag 1978: Mohapatra M, Anand S Synthesis and applications of nano-structured iron oxides/hydroxides- a review. Int J Eng Sci Technol 2:20.